AWS adds a document extraction skill for AI agents to Textract Core

Amazon Web Services added an agent skill to Amazon Textract Core on November 20 that lets AI agents extract text, tables and form data without a team building its own Textract integration. The skill selects the Textract API that fits the document and formats the result, AWS said.
The skill can process multipage PDFs and return their text in reading order. Its parser also turns responses into structured fields, such as invoice details or answers to questions about a form. It uses Textract confidence scores to flag uncertain results so a team can send those fields for review rather than treating every result as equally reliable.
AWS says agents using its MCP server with the aws-core plugin can load the skill automatically. It is also distributed through the Agent Toolkit for AWS, which supports coding agents including Kiro, Claude Code and Cursor. The skill is available in the commercial AWS regions that already offer Textract. There is no separate charge for the skill, but standard Textract usage charges still apply.
The update makes the existing document-reading service easier to use from an agent and gives teams a way to review uncertain fields. It is separate from Textract Insights, the service AWS announced earlier for extracting information using customer-defined document templates.
Sources and citations
Revision history
Published .